Background: Body mass index (BMI) does not distinguish adiposity from lean mass and may therefore inadequately capture heterogeneity in diabetes prevalence. We assessed BMI–DXA obesity concordance, examined diabetes prevalence across DXA-derived body composition phenotypes, and evaluated whether lean mass modifies the adiposity–diabetes association.
Methods: We analyzed 7950 U.S. adults aged 20–59 years from NHANES 2013-2018. DXA-defined obesity was determined using sex-specific optimized-diabetes fat mass index (FMI) thresholds, and low appendicular lean mass index (ALMI) was defined using sex-specific 10th percentile thresholds. Participants were classified according to BMI-DXA obesity concordance and DXA-derived body composition phenotypes. Associations with diabetes were evaluated using survey-weighted logistic regression models.
Results: More than one in ten adults (11.2%) were classified as having obesity by DXA but not by BMI, and this group exhibited higher odds of diabetes (OR 1.91, 95% CI: 1.34–2.72). Individuals with both high FMI and low ALMI exhibited the greatest odds of diabetes (OR 4.49, 95% CI: 2.94–6.56). A significant interaction between high FMI and low ALMI (interaction OR 1.78, 95% CI: 1.25–3.47) supported a modifying role of lean mass in the adiposity-diabetes association. This modifying role differed across BMI categories (P-interaction =0.032) and was most pronounced among individuals with overweight.
Conclusion:
BMI and DXA-derived body composition measures identified distinct obesity classifications associated with different diabetes prevalence and odds. Low lean mass appeared to amplify the adverse metabolic association of adiposity, particularly among individuals with overweight, suggesting that the association between adiposity and diabetes differs according to lean mass status.